Compaq ProLiant 1000 Compaq Backup and Recovery for Microsoft SQL Server 6.X - Page 33

Doc No 444A/0797

Page 33 highlights

Compaq Backup and Recovery for Microsoft SQL Server 6.x Page 33 Device Controller Bus(es) - saturation of Fast-Wide SCSI-2 bus(es) (10/20 MB/sec) on 32-Bit SCSI-2 controller or Smart-2/P Array controller to which tape or disk dump devices interface.48 Dump Device - limited capability of tape or disk device to service write requests (1.25/2.50 MB/sec49 for DLT Tape drives). Intermediate (Link): Network - insufficient bandwidth (1.25/12.5 MB/sec50) in the network cable. NIC Cards - limited capability of network controllers (at server or client) to process I/O requests to/from the network. Backup processes generally involve sequential I/O meaning that successive read (or write) requests are issued for blocks of data that are stored contiguously on disk. Sequential requests can be processed by disk drives in rapid succession, because head seek times are minimal, and so each drive can be made to deliver very high throughput depending on the size of the requests (larger size requests yield higher throughput) and on the spin rate (RPM) of the drives. Sequential reads also allow for the use of 'readahead' algorithms in the hardware (like that in the Smart-2 controller) or in the application. Thus, when planning a strategy for optimal backup performance, it is important to remember that sequential I/O to the disk drives be maintained whenever possible. This concept will be stressed in this paper whenever applicable. Chart-3 represents results of SQL Server dumps to various single tape devices. In this section, the dump devices consist of a 15/30-GB DLT tape drive and a 35/70-GB DLT tape drive. A 10/20-GB DLT tape drive was not tested because its performance is identical to that of the 15/30-GB DLT. SQL Server versions 6.0 and 6.5 were both tested with each of the DLT drives, with Windows NT 4.0 providing the operating system platform.51 48 We have recommended that Wide-Ultra SCSI controllers be used for reasons explained earlier. However, these controllers will operate in Fast-Wide mode with wide SCSI devices, or Fast-SCSI-2 mode with narrow devices. 49 Numbers indicate data transfer rates with compression disabled and with 2:1 compression. 50 Numbers in parenthesis indicate maximum data transfer rates for 10BaseT and 100BaseT ethernet. Actual rate is lower. 51 We have not observed a significant difference in performance between Windows NT versions 4.0 and 3.51 for the purposes of these tape backup tests. 1997 Compaq Computer Corporation, All Rights Reserved Doc No 444A/0797

  • 1
  • 2
  • 3
  • 4
  • 5
  • 6
  • 7
  • 8
  • 9
  • 10
  • 11
  • 12
  • 13
  • 14
  • 15
  • 16
  • 17
  • 18
  • 19
  • 20
  • 21
  • 22
  • 23
  • 24
  • 25
  • 26
  • 27
  • 28
  • 29
  • 30
  • 31
  • 32
  • 33
  • 34
  • 35
  • 36
  • 37
  • 38
  • 39
  • 40
  • 41
  • 42
  • 43
  • 44
  • 45
  • 46
  • 47
  • 48
  • 49
  • 50
  • 51
  • 52
  • 53
  • 54
  • 55
  • 56
  • 57
  • 58
  • 59
  • 60
  • 61
  • 62
  • 63
  • 64
  • 65
  • 66
  • 67
  • 68
  • 69
  • 70
  • 71
  • 72
  • 73
  • 74
  • 75
  • 76
  • 77
  • 78
  • 79
  • 80
  • 81

Compaq Backup and Recovery for Microsoft SQL Server 6.x
Page
33
1997 Compaq Computer Corporation, All Rights Reserved
Doc No 444A/0797
Device Controller Bus(es) - saturation of Fast-Wide SCSI-2 bus(es) (10/20 MB/sec) on 32-Bit
SCSI-2 controller or Smart-2/P Array controller to which tape or disk dump devices interface.
48
Dump Device - limited capability of tape or disk device to service write requests (1.25/2.50
MB/sec
49
for DLT Tape drives).
Intermediate (Link):
Network - insufficient bandwidth (1.25/12.5 MB/sec
50
) in the network cable.
NIC Cards - limited capability of network controllers (at server or client) to process I/O requests
to/from the network.
Backup processes generally involve sequential I/O meaning that successive read (or write) requests are
issued for blocks of data that are stored contiguously on disk.
Sequential requests can be processed by
disk drives in rapid succession, because head seek times are minimal, and so each drive can be made to
deliver very high throughput depending on the size of the requests (larger size requests yield higher
throughput) and on the spin rate (RPM) of the drives.
Sequential reads also allow for the use of ‘read-
ahead’ algorithms in the hardware (like that in the Smart-2 controller) or in the application.
Thus,
when planning a strategy for optimal backup performance, it is important to remember that sequential
I/O to the disk drives be maintained whenever possible.
This concept will be stressed in this paper
whenever applicable.
Chart-3 represents results of SQL Server dumps to various single tape devices.
In this section, the
dump devices consist of a 15/30-GB DLT tape drive and a 35/70-GB DLT tape drive.
A 10/20-GB
DLT tape drive was not tested because its performance is identical to that of the 15/30-GB DLT.
SQL
Server versions 6.0 and 6.5 were both tested with each of the DLT drives, with Windows NT 4.0
providing the operating system platform.
51
48
We have recommended that Wide-Ultra SCSI controllers be used for reasons explained earlier.
However,
these controllers will operate in Fast-Wide mode with wide SCSI devices, or Fast-SCSI-2 mode with narrow
devices.
49
Numbers indicate data transfer rates with compression disabled and with 2:1 compression.
50
Numbers in parenthesis indicate maximum data transfer rates for 10BaseT and 100BaseT ethernet.
Actual
rate is lower.
51
We have not observed a significant difference in performance between Windows NT versions 4.0 and 3.51 for
the purposes of these tape backup tests.